OpenAI launches Astra for Law, pairing GPT-6 Astra with a 230-million-URL legal search index

OpenAI announced Astra for Law on 2026-09-17, available as gpt-6-astra-law: GPT-6 Astra plus a legal index of 230M+ U.S. law URLs built on Free Law Project's CourtListener. It scored 54.0% on Vals AI's Legal Research Bench (200 questions) versus 38.7% for…

OpenAI introduced Astra for Law (announced 2026-09-17), combining its GPT-6 Astra model with a legal search index spanning more than 230 million URLs of U.S. case law, statutes, and regulations. The index was built with Free Law Project's CourtListener and is said to cover 99.9% of U.S. precedential case law. On 200 Vals AI Legal Research Bench validation questions, Astra for Law scored 54.0% overall correctness versus 38.7% for GPT-6 Astra with plain web search; both reports agree on these figures, though The Decoder characterizes the benchmark as a self-run, self-reported test by OpenAI. The launch includes 26 ecosystem plugins (including for Relativity and Clio), a Trusted Access Program offering law firms zero data retention, and API partners Harvey and Legora that can build on the product. The Decoder adds that Anthropic is also expanding its presence in legal AI.

• OpenAI announced Astra for Law on 2026-09-17; it is available as gpt-6-astra-law.
• The legal search index covers 230M+ URLs of U.S. case law, statutes, and regulations, built with Free Law Project's CourtListener and covering 99.9% of U.S. precedential case law.
• Benchmark: 54.0% on 200 Vals AI Legal Research Bench validation questions versus 38.7% for GPT-6 Astra with web search alone (a 15.3-point gain); The Decoder notes the test was self-run/self-reported by OpenAI.
• The offering includes 26 ecosystem plugins, including integrations for Relativity and Clio.
• A Trusted Access Program gives law firms privacy controls including zero data retention.
• API partners Harvey and Legora can build on the product.
• The Decoder reports Anthropic is also expanding its presence in legal AI.